The Rising Role of AI in Healthcare

Transcript generated by AI:

Announcer: 0:00

Welcome to MedEvidence, where we help you navigate the truth behind medical research with unbiased, evidence-proven facts, hosted by cardiologist and top medical researcher, Dr. Michael Koren.

Dr. Erich Schramm: 0:11

Hello, and welcome back to another episode of the MedEvidence Podcast. I'm your host, Dr. Erich Schramm. I'm a board-certified family physician and longtime clinical research investigator with more than 20 years experience with the ENCORE Clinical Research Group. I'm very excited to be talking to Dr. Robert Pearl today. Dr. Pearl is the former CEO of the Kaiser Permanente Group and spent more than 18 years guiding that organization. In addition, he is a board-certified plastic surgeon and reconstructive surgeon. He is on faculty at the Stanford School of Medicine as well as a graduate school of business. So he's well versed as a clinician and in the business of medicine. So in addition, he is a multi-prolifative author and most recently has authored ChatGPT MD, how AI Empowered Patients and Doctors Can Take Back Control of American Medicine. So welcome to the Med Evidence Podcast, Dr. Pearl.

Dr. Robert Pearl: 1:22

It's a privilege to be here. Thank you.

Dr. Erich Schramm: 1:24

Thanks. So to start with, one of the things that comes just about every day in the office, I have patients that come up to me and physicians and say, gosh, why is this medical system so broken? Patients, caregivers, physicians, how did we get here? How did we get to such a broken system?

Dr. Robert Pearl: 1:45

As you note, I've written three books. The first one was on the system of medicine and the problems. The second one on the culture of medicine, the third on technology. And what I see is that we still have retained an approach to medical care left over from the 20th century. You know, in the 20th century most of the problems we took care of were acute medical problems. You came in with a broken bone and the orthopedist fixed it. You came in with a pneumonia and an antibiotic was prescribed that usually was able to solve the difficulty. And yet in the 21st century, we're facing a whole different set of problems. We're seeing chronic disease, we're seeing uh stagnation in life expectancy. We're seeing far more complex patients who need a different type of care than we were able to provide in the past. The American healthcare system, from my perspective, is broken. It's 19th century cottage industry. It's fragmented. Doctors are unconnected with each other, often unaware of what they're doing. They get paid on a piecemeal basis. We call it FIFA service. The more you do, the more you get paid. Whether there's any good or not, doesn't matter. See a patient who you can take care of in one visit and you spend two, you make twice as much money. And we use technology that's still left over from the last century, although in reality it's from the century before. The fax machine, an 1834 invention. I tell my students at the Stanford Graduate School of Business about the fax machine and they ask me, "what's a fax machine?" And so I think the problems of today have just grown across 50 years with a stagnant healthcare system in the context of ever-changing demands, both medically and financially.

Dr. Erich Schramm: 3:35

That's a great point. Just like those residents who may not fully know what a pager was. What is that thing, right? Faxes and pagers. And so, right, so you talk of uh making great points about kind of the disconnect that patients experience in the office or you know, physicians experience and and how that impacts patient, ultimately affects patient uh care and patient outcomes.

Dr. Robert Pearl: 4:03

Well, I think the biggest piece for listeners uh to recognize is how generative AI will now give us a new opportunity to be able to make those changes. As you know, I call the book Chat GPT MD with subtitle at How AI Empowered Patients and Doctors. I start with the patient because medicine still retains, I'll call it a paternalistic view. The patient comes to us, we tell them what's wrong, we tell them what to do, and we give them an appointment to return in the future. Uh, if you look at a patient, let's say, with hypertension, you're a primary care physician, you take care of lots of people with hypertension. You see them in the office, you write a prescription, you diagnose hypertension, you write a prescription, you say I'll see you back in four months. Four months later, the blood pressure is still elevated. You say it's a white coat syndrome. It could be a year before a patient with hypertension is managed. It shouldn't come as any surprise that hypertension accountable for 40% of strokes. We control the United States 50% of the time now. Diabetes, even less often. You know, what we need to have is an approach that is going to be continuous monitoring, so an approach in the home. And I believe that we need to change our relationship to patients and now recognize that whereas in the past maybe they couldn't do as much to not just advance their health and protect themselves, but also to actually be first-line intervention for themselves. You know, we were so afraid of Dr. Google, and that was appropriate because this was simply giving patients 20 links to different problems. They couldn't sort out what it all meant, they couldn't figure out which ones were valid, which ones weren't. The large language models of today, and although I call it ChatGPT, it's just as true about Gemini or Claude, it doesn't matter which LLM you use, they're relatively equivalent. They're very accurate. They give very good results. You know, as you say, I'm a plastic surgeon, I focus on kids with cleft lip and cleft palate. And when I have a problem that's outside of my specialty, I check with Chat GPT first, and then I call a colleague and ask them what's really going on. I'm amazed 99% of the time the advice is identical. We need to build that into the practice of medicine, change our expectations, be able to have patients with, let's say, hypertension, use the technology, combine it with a home blood pressure cuff, measure blood pressures across an entire month. No physician wants a hundred readings per month, but a generative AI tool can analyze progression, first and second derivatives, and if the patient is not getting better, it can notify the patient, and the patient can contact the physician about the need for medication change. And you don't need to see them in the office. You know what's going on, you have the data. We've taken a very complex, highly dangerous problem and simplified it, taking very little of the clinician's time and getting far better control. You know, we've had this view, and it goes back to decades, that cost and quality and access all compete with each other. And it's not true anymore. The best way to lower cost to make care more affordable is to raise quality. And to raise quality, we need to be able to maintain better access, but that's not doesn't necessarily mean driving to the doctor's office more frequently. It means getting more continuous care. And generally, I can do that, taking pressure off of primary care. And I believe primary care will be where the first and biggest impact is going to be.

Dr. Erich Schramm: 7:43

All right. And that resonates to a primary care doctor. So you again, you're kind of checking off all the boxes. I'm nodding here saying, yes, that makes absolute sense. And, you know, to that point, what AI is is what that means to a patient. And you're already kind of alluding to it, which is, you know, being able in real time to understand, you know, maybe certain physiologic processes, because we wear smart watches or, you know, Aura Rings and things like that. You know, so we are recording a lot of information. We know how people's sleep patterns are, we can know their pulse rates, and we can do EKG. So there's a lot of data that's coming up in there. So AI is actually very patient and user-friendly in that regard.

Dr. Robert Pearl: 8:30

Let me throw a concept out to you if that's okay. Uh, to leverage you off your expertise. Something I recently wrote about in Forbes. You know, I predict that what's going to happen next in primary care is that I'll say maybe 30% of what the typical doctor does in the office, which is managing chronic disease, will be done by technology. And that will free up the clinician now, something that the primary care physician has not had for at least a couple of decades, to be able to do two things. Number one, spend more time with the complex, talking about the multi-morbid patients with many diseases that are not well controlled. And that takes time and time that primary care doctors have not had in the office. But number two, that they actually will also use generative AI tool to be able to take on patients that they currently are referring to specialists, not patients who need intervention or procedures, but ones who need more extensive diagnostic, workup, and outpatient management. And they sent it not because they didn't sort of know what to do, they sent it because they didn't have the time to do it. And so they were just trying to survive every day. I think we're going to see a redefinition in a very positive way for primary care physicians in American medicine.

Dr. Erich Schramm: 9:50

Right. And I'd had the pleasure of reading that in your book, by the way. And I I totally agree with that. Because I mean, if and and you've talked a lot about, and we can talk a little bit later about physician burnout, you know, and you said that if you calculated the total daily hourly, you know, hourly work output for a physician would be over 24 hours a day trying to do manage uh chronic disease, health maintenance, you know, acute, there's not enough hours a day to do that. And I think to your point and understanding and what a generative AI is going to be capable of is relieving a lot of that kind of level of decision. I won't say all the decision-making, obviously, but that level, a lot of that level of decision-making and implementation, right, is would take a huge time burden off physicians that could open up a more opportune time. I'm I always enjoy spending my time face to face with patients. And as we all found out during the health uh healthcare 2.0s, you point out, when we had electronic health records became implemented in offices, that ended up kind of one downside to that was it it became very distracting. Physicians spending their time engaging the computer and not the patient, right? And so looking at an option uh with generative AI that brings the opportunity to allow for more time for the physician to really be able to have, you know, meaningful engagement. And my background is in doing chronic disease management. So again, you're you're checking off all the boxes for me. I, as a person understanding generative AI, and much more clearly after having read your book, that I understand now I'm a I'm a lot more comfortable with the notion that, you know, that this is really gonna it's undoubtedly gonna be the future. I don't have any doubt about that. And I'm I'm not worried that it's gonna take my job away, though we should probably talk a little bit about what my future day in the office is gonna look like in another 10 years or so or five years. But what I'm hearing is that relief. So yes, a lot of that work burden, I think 30%, you know, some, you know, you a lot of physicians may spend 50% of their time doing non-clinical, doing non-clinical workflow in the office. So I think 30% would seem pretty conservative. And especially, you know, how quickly, you know, the AI is progressing, I think I think that's conservative, but I'm I'm happy to hear that. So yes, I I wouldn't have any problem uh incorporating generative AI as a kind of a, you know, as kind of like an assistant, you know, as as an option to improve the workflow through the office.

Dr. Robert Pearl: 12:44

Yeah, if you just look at the demands of today versus the ones of the last century. Yeah, if a patient walked into your office, as I say, let's say with a pneumonia, not a severe one, a mild one, uh, you'd see them, you'd prescribe antibiotic, you might see them a couple more times, and hopefully two, three, four weeks later they would be better. And you might not see them again for multiple years, except for routine type care. Now, a patient walks in your office with diabetes. Remember, 30% of Americans have either diabetes or pre-diabetes. Now, what you have to do is you have to schedule, let's say they're 40 years old, you have to schedule 160 more visits because they're going to be seen four times a year for management of their chronic disease in your office for the next 40 years. You put that across 30% of your practice, it's not surprising in any way that clinicians today, particularly in primary care, are feeling totally overwhelmed. We have driven up the work dramatically without recognizing what's happened because it's happened so slowly, but now what it's doing essentially is making it impossible for you to do the type of work that you did in the past when you had more time.

Dr. Erich Schramm: 13:59

Medicine is an inherently top-down, you know, physician directed for everything. So whether it's creating a note for a patient or a handicapped parking permit or authorized, you know, prior authorized, everything is a top-down and an ability for a process that allows physicians to regain the high ground and focusing on disease management and you know having therapeutic useful relationship with patients. Now it doesn't necessarily have to be in the office, right? I'm I really should be saying, well, that I think could also be useful in terms of on a virtual platform as well. So, and we had, and you pointed out, you know, during COVID, you know, we made some really big steps forward because we were really unencumbered and able to provide for virtual care and get reasonably compensated for it. But after all the COVID contingencies and the end of the of that that emergency, a lot of that got rolled back. And so it would be terrific to see that we would be able to continue to maintain, you know, being able to keep access because virtual is is a much more user-friendly access for patients when you're having to manage non-disease or illness that doesn't you don't have to have people in the office for. So so yeah, I I think if we can improve access and maintain access through through virtual platforms, I'd like to see more, you know, look at us. We don't, we should be texting, we could be able to message our patients, you know, emailing and and have better access that way. But it is not typically the case for for a lot of physician offices.

Dr. Robert Pearl: 15:48

One of the challenges that is, you know, I'm a big proponent of telemedicine for a variety of reasons, but it doesn't take very much less time. It takes almost as much time as an in-person visit. And if you have uh, let's say a paddle of 2,400 patients and 30% of them have diabetes or even 20% have diabetes, you know, you're talking about four or five hundred people. And the only way you can manage them is on a calendar basis. Just imagine if every day you knew which of all of your patients were doing well and which ones required intervention. Imagine if you could contact the people who were having problems that were not under good control today and not have to see the other ones because you had a confidence it was being done. Imagine if a lot of the tests that today, as you say, requires your signature could be done electronically. You know, it's interesting. In Utah today, legislation has been passed that allows a generative AI tool to not only monitor patients with chronic disease, but write the prescriptions for them without a physician intervening. And believe it or not, it has a malpractice insurance coverage that has now been uh provided to it, should it make that you probably won't do, an error. This is where this industry is heading, and it will not just take work off the clinician's back, but take the work for which the clinician doesn't add any more value and free the time up to focus on those patients for whom the doctor makes the big difference.

Dr. Erich Schramm: 17:24

Right. And and point to the fact that how much more high yield an office or virtual visit would be, provided that, you know, in advance that AI has facilitated, knowing that the most recent labs have been looked at and and reviewed, the the medications have been reconciled, that the information and and prompts to, you know, does this patient now need to get that colonoscopy scheduled? And and be able so that, and also from the patient standpoint, that they can make sure that their their blood pressure has been been measured, that they've been keeping track, that when they come into the office, maybe if they're having symptoms or other things, that they may have already inputted that so that that can be made aware that, you know, if a patient's going to be seen and they've been having headaches for the last three weeks, and you know, that may be raising some questions like do you do we need to consider doing a neurologic evaluation, or does this patient need an MRI? Or does this patient- so the promptings I think are really would be very helpful, right? So the patient comes in more empowered with information, and then the physician is able to act on that. And like you said earlier, it's like, okay, well, if so much more can be managed on a looking at an expanded role for primary care, then that means a lot less referrals, right? So a lot of that can be managed more so. So as you pointed out, that you'd have, you know, specialists who can be, or even say surgeons who may be at specialty centers that can really focus on what those referrals that really do need to go, do do need to be seen in the specialist office or the hospital versus those that can be done within the primary care realm. And I I totally agree with that.

Dr. Robert Pearl: 19:13

A bsolutely. I mean, take a picture with chronic heart disease or chronic heart failure. What you'd like to really to know is every day, are their ankles more swollen? What's happened with their weight? Could they uh how did they were they able to breathe well when they lied down last night? How many stairs can they climb? I mean, these are the measures that tell you that deterioration is starting. And if you can intervene now, maybe there's something you could do to prevent the crisis. If not, they're gonna be in the ER in three days. Well, you don't want to get these reports every day coming to your office. First of all, you're not gonna be paid for them under the current insurance system. And number two, it's just overwhelming again, it's just more information that overwhelms you. But imagine if you knew which of your, I'll say 100 patients you're following with chronic heart disease were having this problem today. Now you have three patients to call. That's doable. A hundred is not. So rather than again, an episodic measurement, and of course you can add on top of all of that, what do they eat? Do they have how much salt did they take in place? What's going on with them? It's a very different way to manage a practice based upon patient evaluation, not triage, but evaluation, so that you don't spend time taking care of people on a periodic basis. You might see them once a year, but on a periodic basis, we're doing great. And on the other hand, you have the time not only to see the people in whom you make a difference more often, but more importantly, to see them sooner before the crisis develops. And American medicine needs that. According to the CDC, if we could effectively manage chronic disease as well as the best health systems do in the United States today, they deal with a lot of people and high expense. But if we could do it that well, 30 to 50 percent of heart attack strokes, kidney failures, and cancers would be obviated. Now imagine if we had 30 to 50 percent fewer heart attack strokes, kidney failures, and cancers in the United States today, how different things would be. We'd have a trillion dollars of savings. We could now invest in education and prevention and chronic disease management. We'd create a virtuous cycle, whereas today we have a vicious one.

Dr. Erich Schramm: 21:27

No. No, it it it totally uh resonates. And as a as a family doctor who is very much into preventive care, I feel like generative AI is really dialed, can be really dialed in here, mentioning the fact that you can tailor diet plants to patients, a shopping list, recipes. You know, you can uh yeah, you mentioned something guiding people into better exercise habits or setting up uh doing a yoga session or meditation or a number of things. So it isn't always necessarily have to be something that's pharmaceutical driven, but it also can really impact in terms of lifestyle. So all of that being coordinated. And you know, so I think I I I see the uh the potential in that. And I think again, primary care will be, I think, really at the forefront of watching that, watching that play out. So I again totally agree it moves from more reactive into being more proactive in health. And I you're you I'm sold on the fact that that's a much more cost effective and from a physician standpoint, you know, it's more rewarding to seeing people you know keeping people healthy and not seeing them get sick and and seeing uh seeing those outcomes. So I I totally agree with that.

Dr. Robert Pearl: 22:47

You know there's an interesting system that the Mass General Brigham health system, the Harvard-related system, put in place. They call it Care Connects. They faced a situation where they didn't have enough primary care physicians and patients calling in often got appointments months later, which was not acceptable to the kinds of problems they were having. And in this Health Connect system, the interaction is first with the generative AI tool and it takes a history from the patient. So it listens in about all the symptoms it makes a diagnosis and a recommends treatment. And out of all of that information a clinician then schedules as you've noted a telehealth medicine isn't because now they have all the information. They have the timing around it. Mayo Clinic is doing a very similar type of intervention. Cedar Sinai has already done it and of interest what they did is they took patients who were communicating the problems that they were having and they compared the response that the generative AI tool would have done compared to what the clinicians did and they rated the quality of the answers 10% higher for the generative AI tool than for clinicians. Just think about that these tools can do the type of I'll I'll call it routine care. It's not just necessarily chronic disease or predictable care, but it's routine care. These are not the really really sick patients. These are the people who come to our office and they can do it for the patient in the comfort of their home reliably and safely to me it's like it's it's like the ATM machine you know the ATM machine when it first came in everyone was afraid it was going to eat their money and make it disappear.

Dr. Erich Schramm: 24:36

That's right taking their money.

Dr. Robert Pearl: 24:38

That's what we prefer. We don't really want to see a bank tell it because it's just inconvenience or we want to do it at home on our phone. I think it's important for listeners to understand this is not theoretical. These are the things that are happening today and we now should be getting ready for the types of changes you just mentioned where it's going to change the practice of primary care significantly and I'll add this key piece because I'm a big believer that if we don't move from FIFA service to capitation all the good things are just not going to happen. Because the reality is that if we decrease heart attack strokes and kidney failures and cancers a lot of physicians would find their incomes plummeting. The good news is that in a capitated system they would see their incomes rise. And more significantly I think the kind of projected $100,000 clinician deficiency that's being seen for American medicine over the next decade simply won't develop. People aren't going to lose their job they actually will have better jobs, more fulfillment, better income with less work because they will be able to focus on keeping people healthy rather than simply reversing severe disease once it ensues.

Dr. Erich Schramm: 25:55

I and and in your book makes a really compelling you know argument in favor of capitation over fee for service and in in my practice time I spent a good deal of that time in the fee for service model. And so aware of certainly a lot of the shortcomings in that and you you're right and and especially in a model in which you're trying to deal with chronic disease your yeah capitation may make a lot more sense and kind of re-incentive at least in primary care reincentivize doctors and and and I think that's an important point because at the end of the day you've you've got to you know you've got to have you got to have the the the the physicians buying in on that and and all in on that the other interesting thing and you'd mentioned and you did recent programs and and how AI was actually outperforming physicians. And I think also you'd mentioned that there was a study that was measuring like the bedside manner. And it was interesting what you'd pointed out when you know AI regenerative AI versus versus physician do you recall that?

Dr. Robert Pearl: 27:07

Certainly so what you're talking really about two aspects and they're both very important. The first one is the quality of the information and again what we tend to want to look at a generative AI tool actually at any new technology and measure it against perfection. We fail to measure it against what doctors actually do today. We have we know that 4000 Americans die every year from misdiagnosis another 4000 die from develop they don't die but they develop long-term major disabilities and the reason isn't because clinicians don't know what to do it's not because they're not motivated to do it they just don't have the time and so the studies are interesting they compare outcomes driven by a generative AI tool against the clinician and they find the generative AI tool does a better job. Some of which is more comprehensive. But the particular study you're alluding to is from Arizona University where they took uh text and email messages as you mentioned earlier that were given to clinicians and they gave the same messages to a generative AI tool and now they had two sets of responses and they brought in independent observers to measure them. The independent observers didn't know whether they'd been generated by a human or generated by a technology because all they had was the message back to the patient. And they rated the ones generated by the generative AI tool not only as better in terms of comprehensiveness and quality but four times better in terms of empathy. I think that's something to recognize that again in our haste to get through the day to get through the number of patients that we need to see in the 15 minutes we have to see a patient we don't demonstrate the empathy we cut them off earlier. We sort of give them the answers we don't want to hear their whole story and that is something that the tool can provide to patients not just better answers and outcomes but equally greater empathy and support.

Dr. Erich Schramm: 29:16

Well like I said it's it's encouraging because I again that that that because again if physicians are concerned that it's going to be this kind of robotic interface and and that's patients will not you know it's going to turn patients away or give some negative impression that this is this is reassuring. But back to your point if the if you're able to unload a lot of that kind of unnecessary or trivial workflow for physicians it opened up more time then I think that's a big point in terms of taking you know addressing the physician burnout burnout that that's been epidemic. So I feel like those those things are going to be two positives. You'd also uh I was going to look at you had mentioned also these kind of recent advances Chat GPT health I understand that was also recently unveiled what's what's that all about?

Dr. Robert Pearl: 30:10

To me this is a very fascinating development I mean over the past three years if you look at the large language model companies these are OpenAI and Google and Anthropic they have consistently said that the tools they were developing ChatGPT Gemini and Claude were not to be used for the direct revision of medical care. And yet at the same time OpenAI announced that 230 million people every week use their tools for medical information and advice that's when OpenAI released this data repository patients can put in information from the electronic health record into it. It also has the capability as you mentioned earlier of being able to download information using Bluetooth technology off of their smartwatches and potentially other devices they might have. And this is going to be an opportunity in essence to displace potentially the current electronic healthcare record system. Certainly it will be certainly it will allow individuals as they go from various clinicians to bring information with them so that there is a common set of data that all the clinicians can utilize. But to me what's most interesting about it is not creating the data repository. You may remember at least a decade ago Google tried to do the same thing and people weren't that excited about doing so. I think the real opportunity is that this is opening the door to the future. You know I told I think about it like Julius Caesar bringing his troops across the Rubicon once they had marched into the Roman Empire violated a longstanding prohibition against bringing an army into the Roman Empire itself the die was cast and the past would never be returned to because the next thing in the repository is going to go information off of wearable monitors and home monitors like your glucometer at home, your blood pressure cuff measurement maybe your scale a whole bunch of information is going to be downloaded seamlessly on your behalf. Why that is so significant I think is number one in terms of chronic diseases we've mentioned it now will allow the tool on a day-to-day basis to know how you're doing it can inform you when you're not under control. And I agree with you prevention is what we should be doing. The reality is prevention's really hard and we have to also recognize that it's the patients who have longstanding chronic disease who are the ones who are going to develop the complications in terms of the heart attacks and strokes and kidney failures. Number two, it's going to allow you to ask questions of it 24 by seven. And there's another development in medicine happening today there's 170 home testing opportunities now available to patients. 170 different ones and we're not talking about minor things we're talking about diagnosing sexually transmitted diseases. We're talking about the ability to do a cervical cancer screening with a swab at home and the opportunity now when the information comes back to be able to educate the patient what to do this is the dawn of a new area we shouldn't think about it as the end point. That's always the mistake when you have an exponentially growing technology like ChatGPT it's doubling in power every year. In fact the things that I thought when I wrote the book ChatGPT MD three years ago that would take at least five years are already happening. We are moving at a speed of light forward in this technology both in terms of the reliability and quality excellence but I think equally maybe more importantly patient acceptance and use of it. And so I think the time has really come now to move ahead recognize that with these capabilities ask ourselves what do we want to do and I'll throw out another idea and it's going to sound totally radical because again it's so at the cutting edge right now what's happening right now in general technology that I think is going to be most interesting is called vibe coding. If you wanted to create your own application to help take care of our patient you'd have to hire Teta engineers pay them probably $100,000 to write it for you we now have tools the tool from Anthropic the tool from OpenAI that's going to allow people to create these applications very specifically in a short amount of time for people who have no background in information technology have no idea how to code you'll ask it to create it it can create websites that's the easy part now start looking at the possibility that you as a clinician will create a tool specific to your patients that will talk about when the blood pressure gets out of line when the blood sugar gets out of line you'll include within it aspects around lifestyle medicine changes. It'll be specific to how you practice it may even be able to then be able to sign the documents you mentioned earlier if a patient needs to have a placket for their car for a disability to be able to park in spaces closer to a building it probably will allow you to complete the forms for patients who have pediatric vaccination requirements you can go down the list of opportunities all the tasks that you don't want to do you're not going to have to buy this application. Now will it be three years two years four years I don't know exactly what it's going to be but I think it's going to be just as easy as now it is for people to use their phone or the or their computer to be able to access the internet and book a hotel book a flight do all the things that we now do without thinking about in modern in healthcare in in our lives today will be just as relevant in healthcare in the future.

Dr. Erich Schramm: 36:32

You know I radical is not necessarily a bad term. So I think this is a great point to make and understanding again that seeing that physician burden being lessened and empowering patients. And so one question I have and so when I was preparing the podcast and I under I was listening to your book and I said you had a co-author on your book, right? Yeah chat was your co-author on the book and I said you know what I'm going to use chat to create a podcast outline for Dr. Pearl and I thought one important question that says if patients have AI that rivals physician uh physician knowledge what changes in the doctor-patient relationship and this this is a big one because now patients may be walking in the door with with with more knowledge so how how does that look like in the relationship with the patient?

Dr. Robert Pearl: 37:29

Well first let me clarify for listeners I did it at the time because I wanted to explore how far this technology could go uh calling it a I I did call it a co-author but I think I call it a co-author the same way when a medical student does a research project with me. You know I'm their I'll say I'm their mentor uh the way I started writing the book is I downloaded the 12, the 1.2 million words that I had previously written into ChatGPT although at the time that was a very complex thing to do because the technology had not evolved it now would be easier. And I've got forth in various drafts I have a 30-page bibliography that fact checks everything. I found only one major hallucination a expedition to the North Pole that I don't believe ever happened and if it did I want to apologize to ChatGPT there are actually only six pages in there written by ChatGPT and I highlight them and point it out just so readers can figure out what it's going to be. So I think that that was an experiment that I'm glad that I did because as you note this tremendous opportunity today based upon the technology. So you're raising a very important point we think about the doctor and the technology as competing forces. The question we ask is will this technology replace doctors? And I'm a big believer that we need to shift that from asking the question of whether it's going to replace physicians to asking how do we create a partnership? And by a partnership I don't mean with a doctor is superior and the technology is inferior. That would be a mistake. It's the way we think about medicine as a collegial how does a primary care physician work with a specialist to provide the best care to the patient. And I'm not even saying who's the primary care and who's the specialist between the generative AI tool and the clinician what I'm really saying about it is we need to start asking ourselves okay here's how I've practiced for the past 20 years in the in my office in the community now I have this tool how can I use it differently and I would tell people you know just start with the large language model itself. What do you want your patients to do before they call you or before they contact you if you want them just to call your office the first thing they do then you basically say you want to work longer and harder without getting paid anymore because that's what it means.

Dr. Erich Schramm: 39:54

That's exactly what happens right now. That's correct.

Dr. Robert Pearl: 39:58

Exactly that's the traditional model but if you want to say no what I'd like them to do is to go to the technology I want them to explore the problem and if they still have a question to call me or maybe I want to just drop me as you say an email or a text telling me what they did just so I can check it off the box. How do you want to personalize your practice? I'm talking predominantly to the 40 to 50% of physicians who are still in their own practice somewhere in the community. You know it was interesting to me early in this process writing you know when you publish a book you go on a hundred different podcasts and the host of that show she said to me when we were off air she said my husband fell skiing Dr. Pearl you're a big skier he had his arm over his head he slid a hundred feet three months later his shoulder still hurts he can't use it as well as the opposite side what's going on and I said to her I think I know exactly what he has but do me a favor put the information into chat GPT and then call me when you when you know the answer that the technology is going to give you remember this was early in the process she called back five days later she said thank you so much. I did exactly what you said it said he probably had a rotator cuff tear it said he needed an MRI to establish the diagnosis and I should see an orthopedic expert because he almost definitely needs surgery just think about that compared to a Google browser search where you get a lot of links and a lot of information. She said I went to the doctor he said you probably have a rotator cuff tear she nodded. She knew what the rotator cuff was she knew all the details of the muscle she had studied it. He said you need an MRI. She didn't have to question it. She knew exactly why he was doing it and after surgery he said if I had waited three more months I probably could not have reattached the tendon that had been torn off the bone because the muscle would have contracted. That is patient empowerment and I believe it will lead to better outcomes. And if clinicians don't believe it shouldn't change their practice. But if they do believe it they have to start asking themselves what am I going to do differently? How am I going to accomplish it? And as I say I believe at some point they'll have confidence being able to vibe code just the same way they can use their phone today to do a lot of app, to use a lot of apps to be able to accomplish outcomes they otherwise couldn't in the past would have taken a lot of time. That's where I think it needs to go and I think it's going to radically change it in a very positive way I tell people the combination of a dedicated of dedicated clinicians of of a empowered patients and generative AI will achieve results that are going to be exponentially better than any of the three alone.

Dr. Erich Schramm: 42:43

Well and honestly us physicians are looking desperately for that so back in the day the electronic health record did not deliver that but in the case of looking at this with this on the horizon you know physicians are smart enough to know when when a when a good thing is coming down the pike and recognize the the win-win that involves for themselves and their patients with this. My follow-up question to that is talking about how generative AI is going to reshape medical education because as you know from start time you identify potential med student to practicing doctor you're talking about a seven to ten year timeline. And so and you you've written about the traditional medical education that we all received and but that would would have to look very much different with generative AI. So what what do you see the from the medical education standpoint?

Dr. Robert Pearl: 43:40

You're absolutely correct that medical education will have to evolve. And I want to point out two things the first one is that for the first time that I can think of in history, the entering first year medical students will know things that their professors do not know. They will Have expertise their professors don't have. I can't think of a time. I mean, yes, there was an occasional student with a PhD who came in and had a lot of expertise in a very narrow area. But every entering first-year medical student will be very skilled at using a generative AI tool. So you don't have to worry about teaching them about generative AI, which is what the universities are doing right now. You have to actually have about a week for the faculty to get trained on it before you can do much about it. So where is the expertise of the faculty? It's being able to take care of the patient. And this is where I think the blend's going to happen. You know, when we have a case scenario in which a patient comes in with a given set of symptoms and a given set of findings, the question isn't what should the doctor do? It should what the doctor and the generative AI tool do together. And as strange as it might sound, we're actually going to be inventing it in medical schools right now because we've never had this tool until most recently. And that's what I think we need to do. We need to understand how different every patient is. We need to understand how different circumstances, we need to understand difficult conversations, we need to understand the right labs to order and how the patient can understand what's happening. We need to figure out how we can use this to make the care provided in a hospital smoother and better with greater safety and fewer complications. I think this is the big opportunity. And I don't know whether the deans are going to be capable of acknowledging how much they don't know and how this has to be built. But I think this is going to be the time similar to 100 years ago when we totally redesigned medical education to elevate it. We're going to have to completely redesign medical education so that we train the doctors for the practice they're going to have in the future, not in the past. By the time an entry medical student completes, say, you know, four or five or six year residency after four years of medical school, we're talking about a decade, and medical practice will be completely different than it is today. We need to figure out how we can prepare them for this unknown, not teach them how to practice the same way their predecessors did 20 years ago.

Dr. Erich Schramm: 46:24

Now, you practice in a big academic center. Has anybody come to you looking for any guidance on what that medical education is going to look like in the near future?

Dr. Robert Pearl: 46:35

I do speak with a lot of people about the opportunities that are there, and I point them all in the right direction. Again, what I'll tell you is that when I speak to a group of medical students or group of residents, and I ask you, how many of you are using a generative AI tool every day? Every hand goes up. There's not a single person there who didn't consult it last night to get all the information on the particular problem we're going to be discussing, on the particular topic that we're focusing on right now. They will come in there when physicians get ready to do rounds, to take their boards, they're using generative AI tools. And so, yes, it's happening. The big problem is that the institutions are not yet ready to use these tools for a variety of reasons. One is the economic one. As I said, these tools can improve patients by keeping them healthier. That is not an economic imperative. In fact, if anything, it undermines the number of people in the beds. The institutions are going to have to move from an inpatient to an outpatient type setting. Number two, you have a lot of issues with how hard it is to get clinicians, the culture of medicine to make changes. Often you're dealing with unions with a lot of restrictions. But let's just take an example for an inpatient arena. How does nursing provide care to patients? And they do it based upon the order that the clinician wrote, which says, uh check this patient every Q to Q2 to Q4 hours or whatever the time is going to be. What happens in between? We have no idea. It's like with chronic disease. We have no idea what happens between our personal interactions that are there. But what do we know? We know that when patients deteriorate and have a cardiac arrest on a medical surgical floor and then have to be transferred to the ICU, they're totally resuscitated, they get transferred. The mortality of three months is four times higher. What we know is that when patients end up having a major health crisis, it often can be found an hour or two hours earlier. If someone were looking at the information coming off the monitors at the bedside, we know those monitors create a terabyte of data, 10 with 14 zeros, and 97% of it is never looked at. Now imagine if you had a generative AI tool that could encompass all that information and notify the nurse which patients were having problems that seemed to be getting worse rather than getting better. Imagine a tool that could be at the patient's bedside, listening into each clinician as they come by in the morning and interact with the patient, putting the context of the EHR, including within it data offer those monitors. This technology is so much more powerful than 99% of people understand. It can examine the entire medical literature. We're talking about every textbook and every journal article ever created in a matter of seconds. This is a power that we have not yet fully understood and embraced, which is not to say I'm not very aware of the downsides, the risks of uh for security and privacy, some of the biases that get built into a generative AI tool because it's in clinical practice, the potential hallucination, the great dangers. I'm well aware of those problems, but I'm also equally and maybe more so aware of the opportunities, because most of those problems already exist in medicine. They're just being done by humans and by other technological tools like the EHR. And the big opportunity is now to change how we practice, to save lives, to improve access, to improve service, and to dramatically lower cost. Because if people can't afford to pay for their health care, they're simply not going to get it until it becomes a crisis in the ER.

Dr. Erich Schramm: 50:46

That is beautifully stated. Thank you. I know you've been very generous with your time, and I really have appreciated, like I said, the time I've spent being able to listen to your book. And I encourage everybody, every physician out there, this is going to be important reading, you know, in Med 4.0 with generative AI. Be knowledgeable, be equipped with that. And uh Dr. Pearl, was any question that you wished I'd asked that I didn't?

Dr. Robert Pearl: 51:17

No, I think you did a great job of covering a very, very broad topic. I'll tell the listeners that if you have more questions, you go to my website, which is RobertPearlMD.com. If you want more information, you should subscribe to my monthly musings. It's free, it's ad-free, there's no cost. Just go to my website and sign up for it. You'll get information on all these topics every month, the leading changes in advances, or listen to my fixing health care podcast in which we look at uh the developments each month in medical advances. We have tremendous opportunities, and I'm hopeful in the end that the process of change, the transformation that medicine will undergo, it has to undergo because medicine has suddenly become unmanageable and unaffordable. I'm hoping it'll be led by clinicians, because if it is, I believe we will do the right thing for our patients and the right thing for ourselves. And if not, I can guarantee you that technological companies will come in and they'll just simply use it as a means to have fewer doctors, less care, not greater, superior clinical outcomes. This is the classic story of disruptive technology that Clay Christensen has written about. And I can tell you that every industry he analyzed in his excellent book wished that it acted sooner. I think the same is true for medicine. And I encourage all of our listeners at least to give it a try, to see what happens, and to let me know what you find. I appreciate greatly you having me on this show today. Thank you.

Dr. Erich Schramm: 52:44

And anytime you want to come back, we can tackle any of those other topics and do the deep dive. So thank you again and have a great day.

Announcer: 52:52

Thanks for joining the MedEvidence Podcast. To learn more, head over to MedEvidence.com or subscribe to our podcast on your favorite podcast platform.